The State of Database Dump Compression
Every database team compresses their backups. The question is how well. Here's what the industry standard tools achieve on typical database dumps:
| Tool | Typical Ratio on DB Dumps | Notes |
|---|---|---|
| gzip (most common) | ~15–25% | Default for mysqldump, pg_dump, mongodump |
| zstd (modern default) | ~10–15% | SQL Server 2025 added ZSTD natively; pg_dump v16+ supports it |
| xz -9 (max compression) | ~6–10% text SQL, ~10–40% binary | Best generic ratio, but 5–10x slower than zstd |
| LZ4 (speed-focused) | ~20–35% | pgBackRest, Percona PBM use for speed |
| SQL Server MS_XPRESS | ~20–40% | Microsoft's native backup compression |
| SQL Server ZSTD (2025) | ~15–25% | 30–50% better than MS_XPRESS |
Most teams default to gzip or zstd. The adventurous ones use xz -9 for archival — accepting 5–10x slower compression for the best ratio a generic tool can offer.
We wanted to see how much further a domain-aware compressor could go.
Our Results vs Industry SOTA
We tested on real database dumps across five major database formats:
| Format | gzip | zstd | xz -9 (best generic) | Ours | Win vs xz -9 |
|---|---|---|---|---|---|
| MySQL .sql | ~15% | ~11% | 9.49% | 6.03% | +36% |
| PostgreSQL .sql | ~30% | ~20% | 23.49% | 16.16% | +31% |
| MongoDB BSON | ~15% | ~12% | 12.23% | 7.34% | +40% |
| SQL Server .bak | ~40% | ~25% | 10.46% | 10.19% | +2.6% |
| Oracle JSON | ~8% | ~6% | 4.76% | 3.91% | +18% |
We beat xz -9 — the strongest generic compressor — on every format. The biggest wins are on MySQL (+36%), MongoDB BSON (+40%), and PostgreSQL (+31%).
What This Means for Storage Costs
Database backups are retained for weeks, months, or years. The storage costs compound. For a team generating 100 TB of database dumps per day:
| Compressor | Daily Compressed | 30-Day Retention (S3) | Annual Cost |
|---|---|---|---|
| gzip (~20%) | 20 TB | 600 TB | $165,600 |
| zstd (~15%) | 15 TB | 450 TB | $124,200 |
| xz -9 (~10%) | 10 TB | 300 TB | $82,800 |
| Ours (~7%) | 7 TB | 210 TB | $57,960 |
With a 90-day retention policy, the savings triple. With yearly archival, they compound further.
Lossless, Always
All compression is fully lossless. The decompressed dump is bit-identical to the original — every row, every index, every stored procedure. Safe for production restores.
Try It
Compress your database dumps through our API. Upload a .sql, .bson, .bak, or JSON export and get back a compressed archive.
Sign up for free or contact us for enterprise backup pipeline integration.